Should you tell customers they are talking to AI?

Yes. Telling people that they are talking to an AI is the honest default, and it is the safer choice for your business. The EU AI Act also contains transparency duties for AI systems that interact with people. Below you find the reasons, the rules in broad terms, and wording you can copy.

The short answer and the reasons

Say it, and say it early. There are three plain reasons.

• It is honest. A customer who thinks they are writing with a person is being misled, even if every answer is correct. Most owners would not want to be treated that way themselves.
• People find out anyway. An odd answer or a repeated phrase gives it away. Finding out alone feels like being tricked. Being told at the start does not.
• It sets fair expectations. People who know they are talking to an AI write shorter, clearer questions, forgive a misunderstanding more easily, and know they can ask for a person.

You may read claims that customers prefer or reject AI by some percentage. Do not build on such numbers. Your own customers are the ones who count, and you can ask them.

What the rules say in broad terms

The EU AI Act contains transparency duties for AI systems that interact directly with people. In broad terms, people should be informed that they are dealing with an AI system unless this is obvious from the situation. The duties are phased in over time, and the details depend on your role as provider or user of the system.

Other places have their own rules. Some US states have laws on disclosing bots in certain situations. General consumer protection law in many countries also forbids misleading customers, and pretending that software is a human employee can fall under that.

Data protection rules apply as well. If the assistant collects names, phone numbers or other personal data, your privacy notice should cover it.

Rules differ by country and they change. This is general information, not legal advice. For your own case, ask a lawyer or your trade association. The practical advice stays the same in every country: disclose clearly, and you are on the safe side.

How to say it: wording you can copy

Keep it short and friendly. Say what the assistant is, what it can do, and how to reach a person.

Website chat

Hi, I am the AI assistant of Carter Heating. I can answer questions about our services and prices and book a visit for you. If you would like to speak to our team, just tell me.

Phone

Hello, you have reached Carter Heating. I am an AI assistant. I can answer questions and book appointments. To leave a message for the team, say 'message' at any time.

Text or messaging app

Hi, this is the AI assistant of Carter Heating. How can I help? A team member can take over at any time. Just write 'person'.

At handover

I am passing you to Sam from our team now. Sam can see what we have discussed so far, so you do not need to repeat it.

The last one matters as much as the first. Customers should know when the AI stops and a person starts.

Mistakes to avoid

• A human name and photo with no label. 'Anna from support' with a stock photo suggests a person. If you use a name, add 'AI assistant' next to it.
• Hiding it in the small print. A sentence in the privacy policy is not enough. Say it in the first message or the first seconds of a call.
• Fake human signals. Invented typing pauses, made-up stories or 'let me ask my colleague' when no colleague exists.
• Denying it when asked. If a customer asks 'am I talking to a robot?', the answer must be yes.
• No way out. An AI that never hands over makes people angry. Always offer a person or a call back.

Does being open put customers off?

Some owners worry that people will leave as soon as they read the word AI. Some will. For them, the right answer is a quick way to a person, not a disguise. A customer who was misled and noticed is a bigger loss than a customer who chose the phone.

You can measure this in your own business. Look at these numbers for one month:

• How many conversations were started?
• How many ended after the first message?
• How many asked for a person?
• How many ended with a booking or with contact details?

For example, if 50 conversations start and 8 people ask for a person right away, you know the real size of the group that prefers a human. Then make sure those 8 are served fast.

You can also ask customers directly at the end of a conversation: 'Was this helpful?' A simple yes or no tells you more than any general survey.

A short checklist for your setup

• The first message or first sentence says 'AI assistant'.
• The name and picture do not suggest a human.
• Asked directly, the assistant confirms it is an AI.
• A person can be reached, and the customer is told how.
• The handover is announced.
• The privacy notice mentions the assistant and what data it collects.
• You read conversations regularly and correct wrong answers.

When you compare products, ask each provider how disclosure works. This is one of the 10 questions to ask before you buy an AI receptionist.

The 24/7 AI Receptionist from Astranode, in early access, tells people it is an AI assistant, answers only from your approved information, and hands over to a person.